[Asterisk-Users] Offtopic: improving softphone latency on Linux?
Steve Underwood
steveu at coppice.org
Mon Jan 17 16:43:27 MST 2005
Hi Bruno,
Latencies that big should not be due to the softphone. They are often
due to the sound card driver. The driver for the Yamaha sound chip in my
Vaio, for example, has massive latency like you are seeing. There seems
to be no way to configure that driver to stop it. With the good drivers
you can configure the sound card's buffering - small buffers/low
latency, or big buffers/high latency. The Yamaha driver accepts those
commands, but it seems to have no effect on actual operation. You may
find the latency is dramatically different if you try a different sound
card or a different machine.
Regards,
Steve
Bruno Hertz wrote:
>Hi folks
>
>last weekend, I tried Windows Messenger first time and was stunned by
>the little latency it gives. Until now, I've been using softphones on
>Linux exclusively, like iaxcomm, linphone and sjphone, and they all give
>me about 1, at times even 2 secs delay. Whereas Messenger really seems
>to be in the millisec range.
>
>Of course, I'm now curious why there is that difference. Clearly,
>Messenger is more tightly integrated with the OS and accordingly tuned.
>
>So where does this time go? Kernel? Application level? Web searches seem
>to suggest that sound latency generally is a problem on Linux, so I
>tried the low latency kernel from
>http://ccrma.stanford.edu/planetccrma/software/planetccrma.html
>(there are two kernels, actually, where I only got the stable version to
>boot - bleeding edge didn't do on my machine).
>
>Still, that kernel did not really improve things in a noticeable way.
>
>Question hence: did some of you guys experience and investigate this
>same issue? Any recommendations or hints how to make VoIP even more
>enjoyable on linux?
>
>I wouldn't care that much if I was the only affected party, but of
>course whomever I call will also suffer from those delays, so as the
>staunch Linux advocate I've been so far I'd really like to show better
>performance ...
>
>Thanks, Bruno.
>
>
More information about the asterisk-users
mailing list